Motivation:

Owing to the great success of Deep Learning based pattern recognition in this decade, Machine Learning has recently gained much attention from the communication system research community. It offers several benefits for handling algorithm and model deficits. Since Gaussian Frequency Division Multiplexing is non-orthogonal in the frequency domain, equalization becomes quite complex or the performance degrades when using simple approaches. Therefore, we have such a algorithm deficit.

Goal:

The aim of this project is to first do a literature research in order to find a well-suited machine learning based demodulation algorithm for GFDM with a good complexity performance trade-off. For a restriction of the search space, the investigated approaches are limited to Bayesian learning and neural networks. Finally, the chosen algorithm should be implemented and compared to state-of-the-art approaches.

Requirements:

In order to work on this thesis, a successful participation in the lectures Wireless Communications and Communication Technologies is required. Also, a solid understanding in linear algebra, optimization theory and stochastics is needed. Programming skills in Python are advantageous but not necessary.